Excerpt
Table of Contents
1. Introduction
2. A Theoretical Framework
2.1 Asset Prices, Credit and Financial Crises
2.2 Asset Prices and Income Inequality
2.3 Income Inequality and Credit Growth
3. Asset Prices and their Relation to Credit Growth and Financial Risk
3.1 Collateral
3.2 Wealth Effect
3.3 Bank Capital
3.4 Excessive Credit
3.5 Interest Rates
4. Income Inequality and Equity Prices
4.1 Wage Determination
4.2 Capital Income and Capital Gains
5. Inequality as a Cause of Financial Distress? An Assessment of Recent Literature.
6. An Econometric Model: Determinants of Income Inequality
6.1 Data Description and Pretesting
6.2 Estimator Choice and Results
6.3 Robustness Tests
6.4 Implications for the Relationship between Income Inequality and Banking Crises
7. Conclusion
References
Appendix
1. Introduction
The question that has motivated this paper is whether financial crises and income inequality are systematically related. The long rise of inequality in many advanced countries prior to the Great Recession has inspired several authors (e.g. Fitoussi & Saraceno, 2010; Rajan, 2011; Stiglitz, 2009; Stockhammer, 2012) to argue that inequality is a root cause of this crisis. The suppressing effect of inequality on aggregate demand, these authors argue, has prompted many governments to adopt a debt-led growth model, which relies on over-borrowed, over-consuming households. Additionally, households on their own might respond to growing inequality by saving less, or borrowing more, in order to maintain a standard of living that they deem acceptable (Frank, Levine & Dijk, 2010; Kumhof & Rancière, 2010). This view thus sees inequality as a causal factor for rising debt and credit levels. But while debt and credit are the best predictors of financial crises (Jordà, Schularick & Taylor, 2011) the effect of income inequality on debt seems to be too weak to be considered a root cause (see, e.g. Bordo & Meissner, 2012). The co-occurrence of financial crises and periods of rising inequality may thus be caused by a third factor.
One distinctive feature of financial crises is that most of them are preceded by booms and busts in asset prices. In fact, Reinhart and Rogoff (2009, p. 159) document a “trajectory in real housing prices around all the post-World War II banking crises in advanced economies”. They also identify a similar pattern for equity prices, which tend to peak one to three years before the onset of the crisis after building up for four years on average. Similarly, Borio and Lowe (2002) show that asset price deviations from long-run trends are a significant measure of financial distress. At the same time, various types of assets constitute an important source and determinant of income for many households. Especially top incomes are, for reasons explained in this paper, sensitive to movements in equity prices. According to Galbraith (2012), it is also the top of the income distribution that explains the largest part of the increase in income inequality for example in the US since the 1990s. It is thus reasonable to assume that equity price booms, which often are followed by financial crisis, can generate substantial increases in top incomes. Equity prices might hence be an important link between crises and income inequality.
Most empirical investigations of income inequality so far have focused on long-run determinants related to, for example, technological advances and globalisation, policies and institutions as well as education, which might explain trends and cross-country differences in the level of inequality (see, e.g. European Commission, 2011, or Roine, Vlachos & Waldenström, 2009). The association between financial crises and top income shares, however, appears to be a short-run phenomenon with abrupt fluctuations that cannot be explained by these rather time-invariant variables. In contrast, one has to identify factors that have an almost immediate impact on wages, capital income, capital gains or other sources of income, on the one hand, but are also prone to severe fluctuations during times of financial crises, on the other.
Equity prices thus appear to be an obvious choice. First, the stock markets affect several sources of income with little delay: equity investments often pay dividends; they can potentially be resold at a capital gain; and the performance of company stocks might determine the compensation of top executives in financial and non-financial industries. Second, equity prices, and asset prices in general, are an indicator of financial stability due to their systematic and interdependent relation to credit (Mendoza and Terrones, 2008), and they reveal a systematic boom-bust pattern around banking crises (Reinhart & Rogoff, 2009). The goal of this study is hence to examine these two properties of equity prices in more detail and connect them in order to establish a theoretical framework that explains why financial crises are often associated with a preceding rise in income inequality.
Beyond establishing the theoretical foundation, the link between equity prices and top income shares will be empirically tested utilizing a panel of 18 advanced economies between 1913 and 2011. The significant results are then used to analyze the implications for inequality of income in times of crises. To the author’s knowledge, no such study has been conducted to date. Most closely related is maybe the study by Roine, Vlachos and Waldenström (2009) which suggests a causal relationship between stock market capitalization and top income shares over five-year periods. However, the interest of their analysis is different from the one at hand. The authors use stock market capitalization, in addition to bank deposits, as a proxy for the size of the financial sector in order to examine the distributive impact of financial development in the long run. In contrast, this study is concerned with the direct effect of equity prices on top incomes in order to show that this relation links income inequality to financial crises.
The remainder of the paper is structured as follows. Section 2 introduces a theoretical framework that links income inequality and financial crises through equity prices and proposes a set of underlying hypotheses. The first hypotheses, which establish the link between banking crises, credit growth and asset price booms, are discussed in Section 3. Since this link is already well-established in the existing literature, this section only reviews the theoretical foundation and existing empirical findings. Section 4 looks in more detail on the relation between equity prices and income inequality in order to establish a theoretical link. The main argument is that equity prices, directly and indirectly, affect several sources of income and a price boom may thus increase income inequality. Section 5 gives a critical assessment of a possible causal relationship running from income inequality to financial crises based on recent literature. Section 6 presents an empirical model for the link between equity prices and income inequality, describes the dataset and tests the proposed relationship. Thereupon, the implications and limitations of the results are discussed. Section 7 concludes the paper.
2. A Theoretical Framework
Simply stated, the theoretical framework introduced in the following argues that income inequality and financial crises are systematically related because equity price booms, which tend to precede financial crises, generate income disproportionately for the top of the income distribution. As an introduction, two examples might be insightful in order to motivate this core idea. Figure 1 shows the development of equity prices and income inequality in the US around its two most prominent crises in 1929 and 2007. Equity prices, measured as the natural logarithm of the deflated S&P 500 stock index (grey line, left hand scale), exhibit the suggested pattern. Both crises were preceded by a period of rising equity prices, which peaked shortly before their outbreak and collapsed afterwards. Income inequality, measured as the share of pre-tax income received by the top 1 percent of the income distribution (black line, right hand scale), followed as expected. This measure does not include capital gains, because relevant data is scarce and only available for six out of the 18 countries in the dataset. Consequently, capital gains are excluded throughout this paper, unless mentioned explicitly. If included, the boom in the top 1 percent income share is even more pronounced, as can be seen from the dotted line. In any case, the comovement of both series with the stock market cycle point to their sensitivity to equity prices.
Abbildung in dieser Leseprobe nicht enthalten.
Figure 1: Equity Prices and Income Inequality in the United States
Deflated S&P 500 a) (grey, lhs) and the Top 1% Income Share: excluding b) (black, rhs) and including c) (dotted, rhs) capital gains. Vertical lines indicate first year of crisis d)
a) ln of national stock market index (S&P 500), CPI deflated, t-5=0; b) Top 1% pre-tax income share in %, excluding capital gains; c) Top 1% pre-tax income share in %, including capital gains; d) according to Reinhart and Rogoff (2009).
Sources: Top income shares and CPI from Alvaredo, Facundo, Anthony B. Atkinson, Thomas Piketty and Emmanuel Saez, The World Top Incomes Database, http://topincomes.g-mond.parisschoolofeconomics.eu/; S&P500 from Robert Shiller online data.
“The World Top Incomes Database”, which is the source of the income share data throughout this paper, also provides data on the composition of income for a limited amount of countries. The two crises under consideration are covered. The data shows that between 1924 and 1928 the top 1 percent income share increased by 3.3 percentage points from 16.3 to 19.6 percent. Dividends alone contributed 1.7 percentage points, or approximately one half, to this temporary surge. Wages accounted for 1, interest income for 0.5 and rents for 0.1 percentage points. If included, capital gains added another 3.2 percentage points, almost doubling the total increase. Obviously, the top 1 percent were the great beneficiaries of the stock market boom prior to the Great Depression. The markets yielded direct benefits in the form of capital income and capital gains, and they also may have indirectly justified the increases of top wages.
Between 2002 and 2006 the share of total income received by the top 1 percent increased from 15 to 18.1 percent. The contributions of the individual income sources during this time were 1.8 percentage points from wages, 0.7 from dividends, 0.5 from interest income and 0.04 from rents. Capital gains contributed additional 2.89 percentage points. Again, equity market developments seem to have played a significant role providing direct and indirect benefits disproportionately to the top 1 percent, raising income inequality substantially.
Abbildung in dieser Leseprobe nicht enthalten.
Figure 2: Asset Prices as the Link between Financial Crises and Inequality
At least for the two most prominent financial crises, equity prices thus seem to have been an important factor explaining the preceding rise of inequality, and it seems very likely that this relation is characteristic for financial crises in general. Figure 2 schematically depicts the underlying reasoning linking financial crises and income inequality through asset prices. The following sections briefly introduce the depicted links one by one, and the hypotheses that emerge from this framework will be examined in more detail in the following sections of this paper.
2.1 Asset Prices, Credit and Financial Crises
The double-headed arrow between credit growth and asset prices in Figure 2 suggests that there exists a systematically interdependent and reinforcing relation between both (Mendoza & Terrones, 2008). As summarized by Goodhart and Hofmann (2003), asset prices may theoretically be highly correlated with credit growth for at least three reasons. First, assets may serve as collateral, thereby reducing credit constraints during periods of rising valuation. Second, through their wealth effect they may encourage additional borrowing as households see their net wealth increasing, giving financial leeway to larger consumption and investment decisions. Third, banks’ balance sheets also critically depend on asset prices and they influence the institutions’ willingness and capacity to lend. Beyond that, asset prices and the cost of credit are both influenced by the prevailing interest rate as it determines the time value of money. And finally, excessive supply of credit provided e.g. by expansionary monetary policy or capital inflows, may also feed into asset price inflation (Borio, 2006). On the one hand, it suppresses the interest rate, and on the other, it has to be absorbed somewhere in the economy.
As already mentioned, credit growth is, according to Jordà, Schularick and Taylor (2011), the best predictor of financial crises. The framework in Figure 2 consequently regards credit growth as a proxy for the risk of financial distress. Due to the presumably high correlation between credit growth and asset prices there should then also be a strong link between asset prices and crises. According to Reinhart and Rogoff (2009), the bursting of asset bubbles is most commonly associated with banking crises, because of their deteriorating effect on banks’ balance sheets. On the one hand, there is a direct impact on the balance sheets through the valuation of asset holdings; on the other hand, there is an indirect effect through the quality of loans. If, for example, house prices fall, borrowers’ net worth declines and repayment is less likely. In case of a write-off, the asset and equity positions shrink even further. The framework in Figure 2 thus suggests the following:
Hypothesis 1a: Asset prices and credit growth are positively correlated.
Hypothesis 1b: Banking crises are preceded by asset price inflation (property and/or equity).
Both hypotheses already attracted a lot of research. As stated before, Goodhart and Hofmann (2003) provide the theoretical foundation for the relation between asset prices and credit, and they also summarize some empirical evidence that was found in other studies. Their own research suggests that there exists a lot of simultaneity between asset prices and credit. Especially property price inflation is found to be highly correlated with credit growth and even seems to lead it. The correlation between equity price inflation and credit growth is found to be lower and causal impacts weaker.
As regards the relation with banking crises, Borio and Lowe (2002) show that deviations in asset prices and credit from their long-run trend are a significant measure of financial distress. They also stress the dominant role of real estate prices as a determinant of the credit and business cycle and hence recessions and crises. These findings are supported by Reinhart and Rogoff (2009). They identify a systematic pattern of equity and house prices around banking crises in advanced and emerging economies. Both cycles usually peak one to three years prior to crises after building up for circa four years. The magnitudes of the boom phase in property and equity, however, vary between crises. According to the authors, the “Big Five” crises (Spain ’77, Norway ’87, Finland ’91, Sweden ’91 and Japan ’92), on average, saw their house price indices peaking at approximately 130 in t-1 (t-4 = 100, t = year of crisis), an increase of 30 percent within three years. Equity prices peaked at 110 in t-3. In contrast, property and equity prices both exceeded 130 and 120, respectively, in the US one year prior to the Great Recession. For this paper it is important to identify the individual magnitude of the equity boom, rather than looking at asset prices in the aggregate, because equity can be assumed to affect income inequality much more significantly than property, as will be discussed in Section 2.2.
Since the first part of Figure 2 (hypotheses 1a and 1b) is sufficiently established for the purpose of this paper, this study does not intend to add new findings to the existing literature on the relation between asset prices, credit and crisis. Section 3 will rather take a deeper look at the already existing theoretical foundations and existing empirical findings.
2.2 Asset Prices and Income Inequality
The bold arrow from asset prices to inequality represents the impact that assets, equity in particular, may have on income. Many assets constitute a direct source of income because they may pay dividends (capital income) or can be resold (capital gain). Asset holdings, and most importantly stock holdings, are however unevenly distributed across households and account for differently large shares of total income. An increase in their valuation therefore has a direct but asymmetric impact and raises income inequality. Stocks also have an indirect effect on incomes. They may be used as a company performance indicator and hence determine many top executives’ and bankers’ salaries, which have significantly contributed to the rise in income inequality in the US over the last 30 years (Dew-Becker & Gordon, 2008). A great share of top incomes should thus be sensitive to stock performance. The majority of workers, in contrast, receive a wage that can be assumed to be insensitive to the stock markets. Consequently, asset price booms should be accompanied by a rise in many top wages as well as in capital gains and incomes, while leaving the majority rather unaffected. Income inequality should increase significantly.
The size of the effect on inequality however depends on the type of the booming asset. Houses and real estate holdings are usually not as concentrated as financial wealth. In the US, for example, the bottom 90 percent of the wealth distribution held 52 percent of the former, but only 28 percent of the latter in 2007 (Kennickel, 2008). Real estate is also not as liquid as financial wealth and hence cannot be converted into a source of income as easily. Finally, unlike stocks, most assets do not serve as a company performance indicator and thus have no wage-determining function. Financial asset booms, especially in the stock markets, should consequently result in higher inequality than e.g. booms in the housing market. This relation between equity prices and income inequality is thus and important link between inequality and financial crises.
Some studies already examined the relation between financial crises and inequality. Atkinson and Morelli (2011), for instance, look at developments of various measures of inequality before and after periods of different types of economic crises, but do not find a systematic relation. Although the authors do mention that the boom and bust of stock markets around financial crises might first disproportionately benefit and later hurt the rich, they do not further substantiate or test their idea. Similarly, Bordo and Meissner (2012) do not find any significant relation between credit growth, as a measure of risk of financial distress, and income inequality. The framework of the paper at hand suggests, however, that the relation between inequality and crises is not that explicit. Firstly, it is not the crisis per se that is related to inequality but the preceding asset price boom, which raises incomes especially at the top of the distribution. Accordingly, the focus should be on inequality measures that stress the top of the distribution. Secondly, the effect on inequality critically depends on the type of asset that experienced the boom. Crises can be very different in this regard and it is necessary to look at the development of equity prices specifically since they can be assumed to have a much higher impact on income than property prices. Additionally, inequality is not necessarily expected to follow a particular trend around crises, but rather to trace the cycle of the stock market. The second hypotheses should thus emphasize the correlation between top incomes and equity prices, and, as a result of hypothesis 1, predict a certain pattern of income inequality during the run-up to banking crises.
Hypothesis 2a: Equity prices are a major determinant of top incomes. Equity price growth is thus positively associated with the growth of the share of income received by the top 1 percent.
Hypothesis 2b: The share of total income received by the top 1 percent should trace the boom-bust cycle of equity prices around crises.
Figure 3 gives a first indication of the validity of these hypotheses looking at 18 advanced economies (see Table A.1 in the appendix for data coverage). The first graph shows the development of income inequality, measured as the average pre-tax top 1 percent income share, excluding capital gains, and the number of countries in banking crisis. The second graph shows the five-year moving average of changes in the income share (black line) and in the natural logarithm of equity prices (grey line). The proposed correlation between equity prices and inequality is apparent even for these aggregated values and amounts to 0.58. The stock market cycle clearly moves together with changes in the top income share and seems to lead it by one or two years.
To some degree, the cycle also follows the assumed pattern around banking crises. The three most significant periods of crises in the 30s, the late 80s/early 90s and 2008 were all preceded by the typical boom-bust-cycle not only in equity prices but also in terms of inequality. The crises of the late 70s/early 80s, however, show that this is not imperative. The savings and loan crisis in the US between 1984 and 1991, for example, was associated with a boom in property prices. Stock market performance and changes in top incomes were rather muted, indicating that the impact on inequality strongly depends on the type of asset, as already mentioned. In addition, not all equity price booms, most notably the dotcom bubble of the late 90s, end in crises, although the associated growth in inequality was extraordinarily high in most of the countries under consideration (18 percent over the period 1994-1999 on average). Both experiences hint to the validity of the presumption that inequality is related to crises through equity.
Abbildung in dieser Leseprobe nicht enthalten.
Figure 3: Income Inequality, Banking Crises and Equity Prices
Average of 18 advanced economies:
Top 1% income share a) (black) and number of countries in crises b) (grey),
change in real equity prices c) (grey) and change in top 1% income share d) (black)
a) top 1% pre-tax income share in %, excluding capital gains; b) number of countries experiencing a banking crisis; c) 5-year moving average of yearly change in natural logarithm of national stock market indices, CPI deflated; d) 5-year moving average of year-on-year change (in percent) of top 1% pre-tax income share, excluding capital gains.
Sources: Top income shares and CPI from Alvaredo, Facundo, Atkinson, Piketty and Saez, The World Top Incomes Database, http://topincomes.g-mond.parisschoolofeconomics.eu/; S&P500 from Robert Shiller online data; all other indices from datastream; dates of crises as suggested by Reinhart and Rogoff (http://www.reinhartandrogoff.com/data/browse-by-topic/topics/7/); and the author’s calculations.
Figure 4 presents the average development of changes in the top 1 percent income share (again excluding capital gains) and real equity price indices around the first year of the 27 banking crises covered by the underlying dataset. In line with the findings by Reinhart and Rogoff (2009), the stock market tends to build up for three to four years, on average, before it peaks at approximately 130 percent one or two years prior to the crisis. The same pattern evolves for inequality. Top income shares grow at 3.2 percent at their peak, contributing to a total 12.4 percent increase over the five-year period previous to the crisis. The first and second year of the crisis (t and t+1) are associated with minor declines in top income shares, but positive growth rates return with the recovery of the stock market. The increase in income inequality is not reversed and remains at a sustained higher level even after the crash. The boom and the recovery seem to create significant gains for the top 1 percent, while the losses of the bust do not generate proportional losses.
Abbildung in dieser Leseprobe nicht enthalten.
Figure 4: Equity Prices, Income Inequality and Banking Crises
27 episodes a) of banking crises in advanced economies (1914-2008):
average of national stock indices b) (grey, lhs)
and the average change in top 1% income shares c) (black, rhs)
a) t=0 indicates first year of crisis, see Table A.1 for crises included; b) average of natural logarithm of national stock market indices, CPI deflated, t-5=0; c) average year-on-year change (in percent) of top 1% pre-tax income share, excluding capital gains.
Sources: Top income shares and CPI from Alvaredo, Facundo, Atkinson, Piketty and Saez, The World Top Incomes Database, http://topincomes.g-mond.parisschoolofeconomics.eu/; S&P500 from Robert Shiller online data; all other indices from datastream; dates of crises as suggested by Reinhart and Rogoff (http://www.reinhartandrogoff.com/data/browse-by-topic/topics/7/); and the author’s calculations.
Section 4 looks at the relation between inequality and equity prices more closely. It will examine the data in a more disaggregated manner and also disassemble different sources of income to discuss their individual dependence on equity prices. This dependence is obvious for capital gains and capital income. They are particularly erratic around crises and prone to short-term fluctuations, which can be well explained with equity market performance. Other sources of income have a less obvious dependence. Wages, which usually constitute the main source of income and hence define the level of inequality, are not necessarily influenced by equity prices at all. Nevertheless, it seems that their relation has tightened over the last three decades for many top earners (Dew-Becker & Gordon, 2008) in many countries, so that there might still be some marked comovement especially during booms and busts.
While Section 4 lays out the theoretical argument for the proposed relation between equity prices and income inequality, Section 6 presents an empirical model and provides some preliminary evidence. The results are then utilized to discuss the implications for the relationship between inequality and banking crises and the limitations of the findings.
2.3 Income Inequality and Credit Growth
The dotted arrow in Figure 2 indicates the idea that high or rising income inequality might result in higher household debt levels. The argument is that, as a result of rising income inequality, low-income households are forced into more debt as they try to maintain a minimum acceptable standard of living (Kumhof & Rancière, 2010) or their social rank relative to their reference groups (Frank, Levine & Dijk, 2010). In other words, a widening income gap inflicts financial strain on households that are falling behind, because they face increased difficulties financing a life that they deem socially tolerable. In the model presented by Kumhof and Rancière this even leads to systematic defaults and financial distress if incomes do not recover and debt levels turn out to be unsustainable. Some authors even argue that inequality could be a root cause of financial crises (e.g. Rajan, 2011; Fitoussi & Saraceno, 2010; Stockhammer, 2012).
Hypothesis 3: Income inequality increases the risk of financial distress.
While hypotheses 1 and 2 create a framework for the relation between inequality and crises with equity prices as an intermediate factor, hypothesis 3 suggests a direct link. This can be either seen an alternative view to the framework proposed in this paper or as a supporting factor. This depends on which link one wants to stress. Obviously, this paper assumes that the indirect link via equity prices is the main explanation for the inequality-crises-relation. The relevance of a possible direct link is discussed in more detail in Section 5, but not taken into account in the rest of this study.
[...]